The article proposes the Masked Space-Time Hash Encoding (MSTH), a novel method for efficiently reconstructing dynamic 3D scenes from videos. MSTH reduces redundancy in storage and computations by representing a dynamic scene as a weighted combination of a 3D hash encoding and a 4D hash encoding. This method minimizes hash collision rate by avoiding redundant queries and modifications on static areas. As a result, MSTH obtains consistently better results than previous methods with only 20 minutes of training time and 130 MB of memory storage.
Publication date: 26 Oct 2023
Project Page: https://masked-spacetime-hashing.github.io/
Paper: https://arxiv.org/pdf/2310.17527
